A common task for computers is to recognize the content of images and classify them in categories. A self driving car uses its video feed to recognize road signs and traffic lights. Image classification is also used to automatically organize photo collections so that you can efficiently look for photos of dogs for example. It can also help companies to automate manual tasks.
ML6 build multiple such image classification systems. For a insurance company we built a tool to automatically evaluate the damage of a car. We also built a system that verifies whether an installation was correctly installed or not.
You give the computer an image as input and the output of a trained convolutional neural network than gives a probability score that indicates to which category the picture belongs. Because of state of the art transfer learning, we can start building a prototype with a dataset as small as 100 labeled pictures.